The Value of the Source: A Day-Two Reflection From the NPA Conference
by Dr. Prince Baawuah, Head of Medicare Advantage Solutions
On day two of the National PACE Association Annual Conference, I stepped away from the booth for two sessions. Different speakers, different organizations, different slides, but they kept landing on the same two words.
"From the source."
Analytics from the source. Documentation from the source. Cost from the source. The participant, seen from the source. By the end of the afternoon, I had stopped hearing it as a phrase and started hearing it as a design principle.
Session One: Every Participant, Every Encounter, Every Dollar
The team from Seen Health, a PACE program in Los Angeles designed with deep cultural understanding for the families in its community, walked through how they see cost and revenue at the level of a single participant.
They told it through an example—a synthetic participant drawn from real patient stories, though her story will be familiar to anyone who has worked in PACE. After a fall in March, her home care needs rose. Her care team raised her authorized home care from 18 to 34 hours a week, and the care was delivered to plan. Her data moved with her care.
But her funding did not. In the example, after a month of medications, clinic visits, home care, meals at the center, and an ER visit, only about 3% of her monthly funding was left, and her needs had just gone up.
Then they asked the two questions that matter:
Was the new diagnosis captured for CMS risk adjustment? Was her level of care updated with the state?
That is a payment question, a clinical question, and a documentation question, all about one person. One of their slides put it simply:
"Every participant has their own care story."
You can only read hers if her revenue, her encounters, her claims, and her care team's time sit in one view.
Seen Health gets there by joining six sources on participant and date of service: state Medicaid revenue reports, CMS membership reports (the MMR), the EHR, claims and vendor invoices, payroll and staff time, and CRM and enrollment. Revenue is checked against eligibility. Only then does an AI agent sit on top, because, as they put it, an analyst is only as good as what it can see. One of the first jobs they gave that agent: flag missing or inconsistent data at the source.
They closed with a self-check worth bringing back to any PACE leadership team:
The self-check
- Is utilization (visits, meals, rides) captured automatically as the work happens?
- Can you see revenue, encounters, claims, and payroll for one participant in one view?
- Do you know this month's cost per participant before the books close?
- Do finance, operations, and clinical teams use the same numbers for the same metric?
Answer yes to fewer than half? Their warning: the gap grows every quarter.
Session Two: Closing the Loop
The second session, from the analytics and health plan operations leaders at myPlace Health, came at the same idea from the risk adjustment side.
They opened with three questions:
- Can anyone in your organization name your current RAF, and what it was last year?
- Do you know which HCCs CMS actually accepted versus what you submitted?
- Do your providers have clear, measurable risk adjustment goals?
The second one stayed with me. They noted that many fully outsourced organizations have never seen an MAO-004 or a MOR. That is the CMS feedback loop: the record of what was accepted and what was paid. If you have never seen it, you are reconciling against what you think you sent.
Their framing of build versus buy was the cleanest I have heard:
"The question isn't whether to build or buy. It's what you have to own either way."
Whatever the vendor arrangement, the organization still owns its RAF, its accepted HCCs, and its evidence.
Their method makes "from the source" operational:
- Prospective review. Before the visit, look back at prior documentation and flag open suspects.
- Point of care. Gaps surface in the EMR during the visit, and the provider documents to MEAT.
- Concurrent review. Coding and billing review open gaps before the encounter goes out.
- Billing stop-gap. Submission is held until every open gap is reviewed and closed.
Nothing incomplete leaves the building. And between the initial assessment and the 180-day reassessment, a lightweight mini-assessment within 12 weeks of enrollment reconciles labs, medications, consultations, and recent hospitalizations, so the care plan doesn't drift for months.
One line from their appendix says it best: return files only tell you about the last two stages of an encounter's life. Encounters that never became a claim, or that died at the clearinghouse, are silent.
Silent encounters are pages of a participant's story that CMS never reads, and you can only find them from the source.
The Value of the Source
Neither session was really about a tool. Both were about the same discovery: every time data passes from one hand to another, something can get lost, and the further you are from where it started, the harder it is to know what.
My colleague and our CTO, Yubin Park, PhD, has been making this argument for a while. When he released the X12 834 eligibility parser in the open-source hccinfhir package, he described the reality many organizations live with: one vendor loads the raw file into a table, another transforms that table, a third team transforms it again into something "analytics-ready," and reconciling one small number takes a string of meetings across teams. His answer was to touch the primary source and go straight to the end goal, because in risk adjustment, eligibility matters as much as the diagnosis codes.
He followed it with direct translation from the 837 to RAF. His reasoning:
"Every data transformation is lossy."
Convert the 837 to a warehouse format and back, and what you think you captured often isn't there. So check the RAF at the input, as close to the source as possible, then again at submission, and match them up. If they agree, nothing was lost.
Put the two sessions next to Yubin's posts and the pattern is plain:
| What "from the source" looks like | |
|---|---|
| Seen Health | Join care, cost, and revenue on participant and date of service, then flag bad data where it enters |
| myPlace Health | Close the gap at the point of care, and hold the encounter until it is closed |
| Falcon Health | Score the 837 directly, next to the EDI, before it goes to CMS, and read eligibility straight from the 834 |
An Old Idea, Newly Possible
None of this is a new principle. In July 2024, I wrote about applying Quality at the Source (QATS) to Medicare Advantage risk adjustment data submissions: make every level of the submission chain (provider, clearinghouse, plan, vendor, CMS) responsible for its own quality, fix errors where they originate, and choose prevention over detection.
What I heard on day two is that PACE organizations are no longer just agreeing with that idea. They are building it. Seen Health is doing it for cost and revenue. myPlace Health is doing it for documentation and encounter submission. Falcon is doing it for payment: connect the data at the source, detect exceptions early, guide the next action, and preserve the evidence.
The common thread is the participant's story. Yesterday I wrote that every participant has an enrollment story, a payment story, a clinical story, and a submission story, and that too often each team holds only one piece. Day two showed where those pieces are still whole: at the source. The visit that happened, the fall that changed her care, the diagnosis her provider documented that morning.
Every handoff after that is a retelling, and something gets lost in each one. Stay close to the source, and you stay close to her.
What I'm Taking Home
Three questions, borrowed from both sessions, that any PACE team can ask on Monday:
- Can you see one participant whole? Revenue, encounters, claims, care hours, and enrollment, in one view, before the books close.
- Do you know what CMS accepted, not just what you sent? If no one on your team has opened an MAO-004 or a MOR, start there.
- Where does your data get checked: at the source, or at the end? If the first check happens after submission, you are already reconciling losses.
Day one was about what Falcon is. Day two was about where the work has to start. The answer from the people building it, in two separate rooms, was the same.
Start at the source, where the participant's story begins.
Every file. Every participant. Every dollar. Reconciled.
See Falcon in Action
If your team is asking these questions, we'd like to show you what reconciling from the source looks like for your participants, your files, and your payments.
Schedule a conversation with Falcon Health
This is my last reflection from NPA. Next, Falcon heads to the SNP Alliance Annual Fall Forum, October 19-20. If you'll be there, come find us.